/basasuya-yolo3

A Keras implementation of YOLOv3

Primary LanguagePythonMIT LicenseMIT

basasuya-yolo3

license

      

A Keras implementation of YOLOv3 (Tensorflow backend) inspired by allanzelener/YAD2K and qqwweee/keras-yolo3

Use Official Parameter

  1. Download YOLOv3 weights from YOLO website.
  2. Convert the Darknet YOLO model to a Keras model.
  3. Run YOLO detection.
wget https://pjreddie.com/media/files/yolov3.weights
python convert.py yolov3.cfg yolov3.weights model_data/yolo.h5
# yolo.h5 is the netwoerk parameter

Training By Yourself

  1. Generate your own annotation file and class names file.
    One row for one image;
    Row format: image_file_path box1 box2 ... boxN;
    Box format: x_min,y_min,x_max,y_max,class_id (no space).
    For VOC dataset, try python voc_annotation.py
  2. download backend from below and place yolo_weights.h5 in model_data folder https://drive.google.com/drive/u/2/folders/1Qj46bDCBWMbIXm7rPZCIjCVIczxaj2l_
  3. Modify train.py and start training.
    python train.py
    You will get the trained model model_data/my_yolo.h5.

Prepare The Environment

  1. install python3 and some library(recommend through virtualenv)
 pip install -r requirements.txt
  1. install mongodb

Run The System

start three terminal and run the next command.

$ monogod

$ cd Visual
$ python3 manage.py runserver 8000

$ python3 ui.py

Wish you enjoy it!